Tue, 17 Nov 2020 14:43
Open access peer-reviewed chapter
By Yuan Hu
Submitted: April 19th 2016 Reviewed: October 12th 2016 Published: December 14th 2016
DOI: 10.5772/66294
AbstractCurrently, DNA amplification techniques have become important detection tools. However, the extreme sensitivity of such techniques can easily result in contamination. This is a major problem in using these techniques routinely in a regulatory agency such as the Food and Drug Administration (FDA) because false-positive polymerase chain reaction (PCR) results will fail our mission. Preventing PCR carryover contamination and a capacity to rapidly determine false PCR positives are crucial. In the past, several methods have been used to prevent amplicon carryover contamination. In this chapter, we provide practical suggestions for PCR carryover contamination detection and prevention that work well with most PCR applications in our laboratory.
Keywordspolymerase chain reaction (PCR)carryover contaminationnested PCRreal-time PCRsingle-tube nested real-time PCRchapter and author infoAuthorYuan Hu* U.S. Food and Drug Administration, Northeast Regional Laboratory, Microbiological Sciences Branch, Jamaica, NY, USA *Address all correspondence to: yuan.hu@fda.hhs.gov
DOI: 10.5772/66294
From the Edited VolumePolymerase Chain Reaction for Biomedical Applications
Edited by Ali Samadikuchaksaraei
Show +1. IntroductionPolymerase chain reaction (PCR) amplification techniques have provided means for the rapid and sensitive detection of pathogens [1]. The number of applications of PCR is still growing, and more and more amplification-based techniques are now used in FDA field laboratories to detect pathogens, such as Salmonella, Escherichia coli 0157:H7, Shigella, Vibrio, hepatitis A virus (HAV) and noroviruses (NoVs) [2]. A significant challenge facing us is that the sensitivity of PCR can easily result in contamination and consequently in false-positive PCR. A small amount of previously amplified PCR product or potential target sequences that infiltrate laboratory supplies and equipment or that are present in an aerosol can easily contaminate the sample and PCR reagents in the tests. Therefore, prevention of carryover contamination from previous PCR amplifications has become a high priority. As a first line of defense to prevent contamination of PCR with a previously generated amplicon, mechanical separation of the PCR laboratory into different rooms or laboratory benches is needed. Secondly, chemical, UV, and enzymatic methods can be applied to inactivate any prior amplicon generated in the laboratory. Additionally, rapid identification of contaminants and their sources is needed to prevent false-positive PCR results. For this purpose, we developed a rapid method to detect PCR carryover contamination by DNA sequencing. The combination of the above methods plus good laboratory technique should be able to totally eliminate PCR carryover contamination and allow us to perform accurate and sensitive PCR routinely in regulatory setting.
1.1. Polymerase chain reactionIn 1983, Dr. Kary Mullis at Cetus Corporation conceived of polymerase chain reaction. There is not any technique that has had a greater impact on the practice of molecular biology than PCR. PCR-based methods are powerful techniques [3]. This technique is centered around the ability of sense and anti-sense DNA primers to hybridize to a DNA of interest. When put into use, agents of infectious diseases can be detected at extremely low levels. After extension from the primers on the DNA template by DNA polymerase, the reaction is heat-denatured and allowed once again, to anneal with the primers. After another round of extension, a multiplicative increase in DNA products is observed. When critical controls are set, this technique becomes a quantitative process. Therefore, a minute amount of DNA can be efficiently amplified in an exponential fashion to result in an easily manipulable amount of DNA. The current sensitivity and detection limit is at a level as low as 10''50 copies per ml. Although the PCR is extremely easy and fast, PCR product carryover contamination impedes the routine use of these techniques routinely in regulatory laboratories.
1.2. PCR-based technology1.2.1. Reverse transcription PCR (RT-PCR)PCR uses DNA as a starting material. When RT-PCR is carried out, the starting material is RNA. In this method, RNA is first transcribed into complementary DNA (cDNA) by reverse transcriptase from total RNA or messenger RNA (mRNA). The cDNA is then used as a template for the quantitative PCR (qPCR). Real-time quantitative PCR (RT-qPCR) is used in a variety of applications such as food-borne RNA virus and avian flu virus detection. With this technique, we can detect the target RNA at an extremely low level in samples. RT-PCR is an increasingly popular method for RNA virus detection, but DNA contamination in RNA preparations is also a concern. In order to minimize the possibility of carryover contamination in RT-PCR, it is critical to minimize the number of handling and pipetting steps.
1.2.2. Real-time quantitative PCR (qPCR)Traditional detection of amplified PCR product relies upon gel electrophoresis. qPCR is an advanced form of the traditional PCR. It is a major development in PCR technology that enables the reliable detection and measurement of products generated during every cycle of the PCR process. This technique became possible after the introduction of an oligonucleotide probe that was designed to hybridize within the target sequence. Due to the 5'² nuclease activity of Taq polymerase, amplification of the target-specific product can be detected through cleavage of the probe during PCR. These assays are very sensitive and can detect as few as 10''100 viral copies per reaction. qPCR techniques have evolved into a variety of other branches including real-time PCR by Taqman (Roche), LightCycler by (Roche), SmartCycler by (Cepheid), etc. Some of them are now widely in use for virus and bacteria detection in regulatory laboratories. Unlike other PCR methods, qPCR does not require post-PCR product handling, preventing potential PCR product carryover contamination.
1.3. PCR contaminationAll the PCR methods are powerful techniques. Unfortunately, the exquisite sensitivity of these techniques makes them vulnerable to contamination [4, 5]. One of the most important rules when performing PCR is to avoid contamination. This chapter will outline necessary precautions to prevent contamination as well as procedures for detecting and cleaning suspected contamination.
2. Potential sources of contamination2.1. Cross contamination between samplesA large number of target organisms in sample handling may lead to pre-amplification sample cross contamination [6, 7]. The sources of contaminants between samples are diverse and can all contribute to the contamination of the finished PCR product. These sources may include reagents, disposable supplies, sample carryover, improper handling procedures, etc.
2.2. Cross contamination between nucleic acidsCross contamination between nucleic acids is a major problem in all PCR laboratories. Nucleic acids from organisms or plasmid clones derived from organisms that have been previously analyzed and that may be present in large numbers in the laboratory environment could be a source of contamination. Contaminants can also be introduced by unrelated activities in neighboring laboratories. These sources of contamination are problematic as they may lead to pre-amplification cross contamination [8''10].
2.3. PCR product carryover contaminationThe most important source of contamination is from the repeated amplification of the same target sequence, which leads to accumulation of amplification products in the laboratory environment. Even minute amounts of carryover can lead to false-positive results. A typical PCR generates theoretically as many as 108 copies of target sequence [11]. If uncontrolled, amplification products will contaminate laboratory reagents, equipment, and ventilation systems. Carryover of previously accumulated amplified DNA is considered the major source of contamination.
3. Methods to control contaminationContamination between samples and from previous PCR amplicon generation is a significant potential source of invalid PCR results [12]. The first two forms of contamination described above can be easily avoided by using careful technique and good quality control practices. Generally, most PCR-based assays consist of three steps: DNA sample processing, PCR amplification, and amplification product detection (excluding real-time PCR). It is in the latter step that carryover contamination often occurs through methods that include gel electrophoresis, solid phase hybridization, solution hybridization, and capillary electrophoresis [7]. Methods to prevent amplification product carryover contamination have been developed in the past ten years [13''17]. Basically, there are mechanical, chemical, UV light irradiation, and enzymatic methods and closed-tube PCR detection formats, all of which can help to prevent amplification product carryover contamination [7]. The following section will focus on more recent practices and methods that have been used in our laboratory to eliminate carryover contamination.
3.1. Mechanical methodOur laboratory was designed and operated in a way that prevents contamination of reactions with PCR products from previous assays and cross contamination between samples. It includes the separation of areas of the laboratory where samples and reagents are prepared from the areas where amplification is performed and amplification products are analyzed. This unidirectional workflow can reduce the opportunity for contamination to occur. A typical PCR laboratory should be divided into at least three to four different areas'--(1) sample preparation, (2) PCR mix preparation, (3) PCR product detection, and (4) RNase free area'--if the PCR method involves RNA sample.
3.2. Chemical methodGeneral cleaning practices are important for controlling PCR carryover contamination. All surfaces in the PCR area should be routinely decontaminated to prevent cross contamination. The PCR work bench is required to be cleaned with 10''15'% sodium hypochlorite solution (bleach), followed by removal of the bleach with 70'% ethanol.
3.3. UV irradiation methodUV irradiation is an easy method to inactivate amplification product involved in carryover contamination. The method is based on the ability of UV light to induce thymidine dimer formation in the DNA that makes the contaminating nucleic acid inactive as a template for further amplification (Figure 1). A good practice is to expose all of the PCR supplies to UV light for 5''20 min as the nucleic acid will be damaged by absorbing the UV light energy at 254 nm wavelength [19]. UV irradiation is an integral feature of our PCR laboratory, and the Spectrolinker XL-1500 (Spectronics Corporation, Westbury, NY) is used to eliminate contamination that may occur during PCR tests. All of our PCR tools are stored in a UV light box (C.B.S Scientific, Co. Del Mar, Ca). PCR master mix preparation and specimen setup are also carried out in this UV light box.
Figure 1.Action of UV light on the nucleic acids [18].
3.4. Enzymatic methodUracil-DNA glycosylase (UNG) is a DNA repair enzyme [20] that can recognize and remove uracil residues from DNA (Figure 2). In 1990, the use of UNG to inactivate PCR products was first reported [21]. This method employs uracil (dUTP) instead of thymine (dTTP) during PCR to generate amplification products with distinguishing characteristics relative to the native DNA template. Because the newly synthesized amplicons contain dUTP, they are susceptible to hydrolysis by UNG. This method is the most widely used contamination control technique in our laboratory.
Figure 2.Replace dTTP with dUTP during PCR amplification and the PCR product will contain uracil. Prior to PCR, the PCR mixture is treated with uracil-DNA glycosylase (UNG). During the denaturation step, temperature is elevated to 95°C, resulting in cleavage of apyrimidinic sites and fragmentation of carryover DNA. As the template contains thymidine, it will not be affected by the UNG treatment (source: Sopachem Life Sciences).
Briefly, this carryover prevention technique consists of three steps:
The dUTP is incorporated into all PCR products, substituting dUTP for dTTP or incorporating dUTP during synthesis of the primers [15, 22].
Before PCR, mixtures are treated with UNG (Applied Biosystems, Foster City, CA) at room temperature for 10 min to hydrolyze and remove any contaminating amplification products that may be present in the PCR mixtures. This technique also has a hot start function [15].
UNG is thermally inactivated at 95°C for 5 min prior to the actual PCR.
3.5. Another amplification format without the risk of carryover contamination3.5.1. Real-time PCR-based technology to avoid contaminationIn traditional PCR, amplification and detection of the target DNA sequence occur separately. To determine if a sample contains the target sequence, post-amplification handling of the amplicon is required. A more recent technological development, real-time PCR [23], allows for the simultaneous amplification and detection of a target sequence through the use of fluorescent-labeled probes (Figure 3). In comparison to conventional PCR, real-time PCR can reduce the chance of carryover contamination. The new generation of amplification technology simultaneously amplifies and detects target DNA without exposing the amplification products to the laboratory environment. Currently, we developed several real-time PCR methods, such as single-tube real-time PCR and single-tube nested real-time PCR (Figure 4) to simultaneously detect multiple pathogens in a closed system which has substantially reduced the possibility of false-positive results due to amplification product carryover contamination [24, 25].
Figure 3.A typical result of graphical view from our laboratory.
Figure 4.Single closed-tube nested real-time-PCR system: in order to reduce the chance of carryover contamination, all reactions including reverse transcription, conventional PCR, first PCR, nested PCR, and real-time TaqMan detection are performed in a single closed tube [24].
4. Method to detect contaminationIn the context of this discussion, contamination is defined as the unwanted presence of a PCR amplicon. At times, PCR contamination is present but difficult to ascertain. If a significant contamination problem appears in a PCR laboratory, we need to walk through the procedure of testing for contamination and, if necessary, replace all reagents. The PCR parameters considered for potential sources of contamination include amplification setup, amplification product handling, and DNA aerosol and storage. Carryover contamination is determined by the following methods in our laboratory.
4.1. Internal controlsAppropriate control reactions are helpful in determining whether DNA contamination has occurred. It is important to use a special PCR-positive control which is different from the sample DNA, such as a DNA fragment with a deletion or base alteration in the region of amplification [26]. PCR products can be assessed on a gel to distinguish the control from the native PCR products. Negative controls are also very important and must be included with each run because the first sign of contamination trouble is usually the appearance of an amplification product in the negative or blank controls [27].
4.2. DNA sequencingTechniques to sequence PCR products were developed in our laboratory in the past few years [15, 28]. This confirmatory sequencing ensures that the PCR product has the expected sequence. The direct comparison of PCR product sequences from a sample and a control is the best way to determine whether two PCR products are similar or different. After comparison of the DNA sequence variation between the PCR products and the control, the cross contamination of samples can be detected. In some suspected cases, we directly sequenced PCR products by using the ABI BigDye Terminator Cycle Sequencing kit with a 3500 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA) according to the manufacturer's instructions. Briefly, each cycle consists of denaturation at 96°C for 10 s, annealing at 50°C for 5 s, and extension at 60°C for 4 min. After 25 cycles, the fluorescent extension products are purified by a simple isopropanol precipitation step. Software is available on websites [29] to perform a wide range of different types of sequence alignment. DNA sequence data were analyzed by the Geneious, the GenBank sequence database, and the BLAST program from the National Center for Biotechnology Information (NCBI). Accurate identification of any contamination is required for the proper function of FDA field laboratory. The DNA sequencing method should be an ideal technology for this purpose.
5. DiscussionPCR is a very powerful and extremely sensitive amplification technique, but there is always the peril that a tiny amount of contamination of the DNA target may lead to false-positive results. It has become necessary to systematically address the issue of PCR contamination, especially in the FDA, a regulatory agency. To overcome this issue, effective methods have been successfully developed and used to avoid carryover contamination in our regulatory laboratories in the past few years [15, 24, 25].
We have effectively established and maintained a unidirectional workflow from a PCR clean to a PCR dirty area, thereby reducing the opportunity of contamination to occur.
Samples were set up on a bench that was isolated from PCR product testing areas.
All PCR master mixes were prepared in a separate room or at least on a separate bench. Also, we always used a separate laboratory coat, gloves, tubes, and filter pipette tips in the different PCR working areas.
A separate aliquot of water stock for each round of PCR was addressed.
All PCR work benches were decontaminated with 10''15'% bleach and 70'% alcohol. All the pipettes, pipette tips, tubes, racks, and gloves were UV-irradiated.
A different pipette tip was used when pipetting each of the PCR reagents, even the same master mix to each tube.
The PCR tubes were kept closed during the procedure. The tubes were opened only when necessary because of potential aerosols that are dangerous with respect to contamination. Minimizing the number of pipetting and mixing steps in PCR master mix preparation is also very important from the perspective of aerosol contamination.
It is very important to schedule PCR when not handling plasmids to prevent cross contamination.
dUTP was incorporated into all PCR products which can subsequently be selectively destroyed by UNG.
Optimization of PCRs is also important. G + C-rich products may be more difficult to inactivate by UNG because of the lower concentration of uridine triphosphate(UTP).
Positive controls consisted of a low copy number of the desired nucleic acid target and should never be prepared or stored with the samples.
Other amplification methods, such as real-time PCR [23] or closed-tube PCR [25] which can reduce the chances of carryover contamination, are now being used more routinely in our laboratory.
A rapid DNA sequencing method to precisely detect contamination was established.
6. ConclusionStandard precautions should always be employed during all PCR-based testing, whether it is real-time PCR or conventional PCR. All the regulatory laboratories should have their own appropriate controls and systematic measures to prevent and detect contamination. When contamination does occur, we need to accurately determine which reagent is contaminated. All of us should also understand that our individual working habits directly affect our work quality. We believe all the above methods can reduce the risk of contamination and ensure the efficacy of all PCR results.
AcknowledgmentsNo official support or endorsement of this article by the Food and Drug Administration is intended or should be inferred.
(C) 2016 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution 3.0 License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
How to cite and referencechapter statistics2095 total chapter downloads
2 Crossref citations
More statistics for editors and authorsLogin to your personal dashboard for more detailed statistics on your publications.
Access personal reportingRelated ContentThis BookNext chapterMultiplex Polymerase Chain Reaction Assay for Early Diagnosis of Viral InfectionBy Hiroko Tsunemine, Yuriko Yoshioka, Miho Nagao, Yasuhiro Tomaru, Toshiharu Saitoh, Souichi Adachi, Norio Shimizu and Takayuki Takahashi
Related BookBioinformatics Tools for Detection and Clinical Interpretation of Genomic VariationsEdited by Ali Samadikuchaksaraei
First chapterThe Bioinformatics Tools for Discovery of Genetic Diversity by Means of Elastic Net and Hurst ExponentBy Leila Maria Ferreira, Thelma Sfadi, Tesfahun Alemu Setotaw and Juliano Lino Ferreira
We are IntechOpen, the world's leading publisher of Open Access books. Built by scientists, for scientists. Our readership spans scientists, professors, researchers, librarians, and students, as well as business professionals. We share our knowledge and peer-reveiwed research papers with libraries, scientific and engineering societies, and also work with corporate R&D departments and government entities.
More About Us\n
1.2. PCR-based technology\n
\n
1.2.1. Reverse transcription PCR (RT-PCR)\n
PCR uses DNA as a starting material. When RT-PCR is carried out, the starting material is RNA. In this method, RNA is first transcribed into complementary DNA (cDNA) by reverse transcriptase from total RNA or messenger RNA (mRNA). The cDNA is then used as a template for the quantitative PCR (qPCR). Real-time quantitative PCR (RT-qPCR) is used in a variety of applications such as food-borne RNA virus and avian flu virus detection. With this technique, we can detect the target RNA at an extremely low level in samples. RT-PCR is an increasingly popular method for RNA virus detection, but DNA contamination in RNA preparations is also a concern. In order to minimize the possibility of carryover contamination in RT-PCR, it is critical to minimize the number of handling and pipetting steps.
\n
\n
\n
1.2.2. Real-time quantitative PCR (qPCR)\n
Traditional detection of amplified PCR product relies upon gel electrophoresis. qPCR is an advanced form of the traditional PCR. It is a major development in PCR technology that enables the reliable detection and measurement of products generated during every cycle of the PCR process. This technique became possible after the introduction of an oligonucleotide probe that was designed to hybridize within the target sequence. Due to the 5'² nuclease activity of Taq polymerase, amplification of the target-specific product can be detected through cleavage of the probe during PCR. These assays are very sensitive and can detect as few as 10''100 viral copies per reaction. qPCR techniques have evolved into a variety of other branches including real-time PCR by Taqman (Roche), LightCycler by (Roche), SmartCycler by (Cepheid), etc. Some of them are now widely in use for virus and bacteria detection in regulatory laboratories. Unlike other PCR methods, qPCR does not require post-PCR product handling, preventing potential PCR product carryover contamination.
\n
\n
\n
1.3. PCR contamination\n
All the PCR methods are powerful techniques. Unfortunately, the exquisite sensitivity of these techniques makes them vulnerable to contamination [4, 5]. One of the most important rules when performing PCR is to avoid contamination. This chapter will outline necessary precautions to prevent contamination as well as procedures for detecting and cleaning suspected contamination.
\n
\n
2. Potential sources of contamination\n
\n
2.1. Cross contamination between samples\n
A large number of target organisms in sample handling may lead to pre-amplification sample cross contamination [6, 7]. The sources of contaminants between samples are diverse and can all contribute to the contamination of the finished PCR product. These sources may include reagents, disposable supplies, sample carryover, improper handling procedures, etc.
\n
\n
\n
2.2. Cross contamination between nucleic acids\n
Cross contamination between nucleic acids is a major problem in all PCR laboratories. Nucleic acids from organisms or plasmid clones derived from organisms that have been previously analyzed and that may be present in large numbers in the laboratory environment could be a source of contamination. Contaminants can also be introduced by unrelated activities in neighboring laboratories. These sources of contamination are problematic as they may lead to pre-amplification cross contamination [8''10].
\n
\n
\n
2.3. PCR product carryover contamination\n
The most important source of contamination is from the repeated amplification of the same target sequence, which leads to accumulation of amplification products in the laboratory environment. Even minute amounts of carryover can lead to false-positive results. A typical PCR generates theoretically as many as 108 copies of target sequence [11]. If uncontrolled, amplification products will contaminate laboratory reagents, equipment, and ventilation systems. Carryover of previously accumulated amplified DNA is considered the major source of contamination.
\n
\n
\n
3. Methods to control contamination\n
Contamination between samples and from previous PCR amplicon generation is a significant potential source of invalid PCR results [12]. The first two forms of contamination described above can be easily avoided by using careful technique and good quality control practices. Generally, most PCR-based assays consist of three steps: DNA sample processing, PCR amplification, and amplification product detection (excluding real-time PCR). It is in the latter step that carryover contamination often occurs through methods that include gel electrophoresis, solid phase hybridization, solution hybridization, and capillary electrophoresis [7]. Methods to prevent amplification product carryover contamination have been developed in the past ten years [13''17]. Basically, there are mechanical, chemical, UV light irradiation, and enzymatic methods and closed-tube PCR detection formats, all of which can help to prevent amplification product carryover contamination [7]. The following section will focus on more recent practices and methods that have been used in our laboratory to eliminate carryover contamination.
\n
\n
3.1. Mechanical method\n
Our laboratory was designed and operated in a way that prevents contamination of reactions with PCR products from previous assays and cross contamination between samples. It includes the separation of areas of the laboratory where samples and reagents are prepared from the areas where amplification is performed and amplification products are analyzed. This unidirectional workflow can reduce the opportunity for contamination to occur. A typical PCR laboratory should be divided into at least three to four different areas'--(1) sample preparation, (2) PCR mix preparation, (3) PCR product detection, and (4) RNase free area'--if the PCR method involves RNA sample.
\n
\n
\n
3.2. Chemical method\n
General cleaning practices are important for controlling PCR carryover contamination. All surfaces in the PCR area should be routinely decontaminated to prevent cross contamination. The PCR work bench is required to be cleaned with 10''15'% sodium hypochlorite solution (bleach), followed by removal of the bleach with 70'% ethanol.
\n
\n
\n
3.3. UV irradiation method\n
UV irradiation is an easy method to inactivate amplification product involved in carryover contamination. The method is based on the ability of UV light to induce thymidine dimer formation in the DNA that makes the contaminating nucleic acid inactive as a template for further amplification (Figure 1). A good practice is to expose all of the PCR supplies to UV light for 5''20 min as the nucleic acid will be damaged by absorbing the UV light energy at 254 nm wavelength [19]. UV irradiation is an integral feature of our PCR laboratory, and the Spectrolinker XL-1500 (Spectronics Corporation, Westbury, NY) is used to eliminate contamination that may occur during PCR tests. All of our PCR tools are stored in a UV light box (C.B.S Scientific, Co. Del Mar, Ca). PCR master mix preparation and specimen setup are also carried out in this UV light box.
\n
Figure 1.Action of UV light on the nucleic acids [18].
\n
\n
\n
3.4. Enzymatic method\n
Uracil-DNA glycosylase (UNG) is a DNA repair enzyme [20] that can recognize and remove uracil residues from DNA (Figure 2). In 1990, the use of UNG to inactivate PCR products was first reported [21]. This method employs uracil (dUTP) instead of thymine (dTTP) during PCR to generate amplification products with distinguishing characteristics relative to the native DNA template. Because the newly synthesized amplicons contain dUTP, they are susceptible to hydrolysis by UNG. This method is the most widely used contamination control technique in our laboratory.
\n
Figure 2.Replace dTTP with dUTP during PCR amplification and the PCR product will contain uracil. Prior to PCR, the PCR mixture is treated with uracil-DNA glycosylase (UNG). During the denaturation step, temperature is elevated to 95°C, resulting in cleavage of apyrimidinic sites and fragmentation of carryover DNA. As the template contains thymidine, it will not be affected by the UNG treatment (source: Sopachem Life Sciences).
\n
Briefly, this carryover prevention technique consists of three steps:
The dUTP is incorporated into all PCR products, substituting dUTP for dTTP or incorporating dUTP during synthesis of the primers [15, 22].
Before PCR, mixtures are treated with UNG (Applied Biosystems, Foster City, CA) at room temperature for 10 min to hydrolyze and remove any contaminating amplification products that may be present in the PCR mixtures. This technique also has a hot start function [15].
UNG is thermally inactivated at 95°C for 5 min prior to the actual PCR.
\n
\n
\n
3.5. Another amplification format without the risk of carryover contamination\n
\n
3.5.1. Real-time PCR-based technology to avoid contamination\n
In traditional PCR, amplification and detection of the target DNA sequence occur separately. To determine if a sample contains the target sequence, post-amplification handling of the amplicon is required. A more recent technological development, real-time PCR [23], allows for the simultaneous amplification and detection of a target sequence through the use of fluorescent-labeled probes (Figure 3). In comparison to conventional PCR, real-time PCR can reduce the chance of carryover contamination. The new generation of amplification technology simultaneously amplifies and detects target DNA without exposing the amplification products to the laboratory environment. Currently, we developed several real-time PCR methods, such as single-tube real-time PCR and single-tube nested real-time PCR (Figure 4) to simultaneously detect multiple pathogens in a closed system which has substantially reduced the possibility of false-positive results due to amplification product carryover contamination [24, 25].
\n
Figure 3.A typical result of graphical view from our laboratory.
\n
Figure 4.Single closed-tube nested real-time-PCR system: in order to reduce the chance of carryover contamination, all reactions including reverse transcription, conventional PCR, first PCR, nested PCR, and real-time TaqMan detection are performed in a single closed tube [24].
\n
\n
\n
\n
4. Method to detect contamination\n
In the context of this discussion, contamination is defined as the unwanted presence of a PCR amplicon. At times, PCR contamination is present but difficult to ascertain. If a significant contamination problem appears in a PCR laboratory, we need to walk through the procedure of testing for contamination and, if necessary, replace all reagents. The PCR parameters considered for potential sources of contamination include amplification setup, amplification product handling, and DNA aerosol and storage. Carryover contamination is determined by the following methods in our laboratory.
\n
\n
4.1. Internal controls\n
Appropriate control reactions are helpful in determining whether DNA contamination has occurred. It is important to use a special PCR-positive control which is different from the sample DNA, such as a DNA fragment with a deletion or base alteration in the region of amplification [26]. PCR products can be assessed on a gel to distinguish the control from the native PCR products. Negative controls are also very important and must be included with each run because the first sign of contamination trouble is usually the appearance of an amplification product in the negative or blank controls [27].
\n
\n
\n
4.2. DNA sequencing\n
Techniques to sequence PCR products were developed in our laboratory in the past few years [15, 28]. This confirmatory sequencing ensures that the PCR product has the expected sequence. The direct comparison of PCR product sequences from a sample and a control is the best way to determine whether two PCR products are similar or different. After comparison of the DNA sequence variation between the PCR products and the control, the cross contamination of samples can be detected. In some suspected cases, we directly sequenced PCR products by using the ABI BigDye Terminator Cycle Sequencing kit with a 3500 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA) according to the manufacturer's instructions. Briefly, each cycle consists of denaturation at 96°C for 10 s, annealing at 50°C for 5 s, and extension at 60°C for 4 min. After 25 cycles, the fluorescent extension products are purified by a simple isopropanol precipitation step. Software is available on websites [29] to perform a wide range of different types of sequence alignment. DNA sequence data were analyzed by the Geneious, the GenBank sequence database, and the BLAST program from the National Center for Biotechnology Information (NCBI). Accurate identification of any contamination is required for the proper function of FDA field laboratory. The DNA sequencing method should be an ideal technology for this purpose.
\n
\n
\n
5. Discussion\n
PCR is a very powerful and extremely sensitive amplification technique, but there is always the peril that a tiny amount of contamination of the DNA target may lead to false-positive results. It has become necessary to systematically address the issue of PCR contamination, especially in the FDA, a regulatory agency. To overcome this issue, effective methods have been successfully developed and used to avoid carryover contamination in our regulatory laboratories in the past few years [15, 24, 25].
We have effectively established and maintained a unidirectional workflow from a PCR clean to a PCR dirty area, thereby reducing the opportunity of contamination to occur.
Samples were set up on a bench that was isolated from PCR product testing areas.
All PCR master mixes were prepared in a separate room or at least on a separate bench. Also, we always used a separate laboratory coat, gloves, tubes, and filter pipette tips in the different PCR working areas.
A separate aliquot of water stock for each round of PCR was addressed.
All PCR work benches were decontaminated with 10''15'% bleach and 70'% alcohol. All the pipettes, pipette tips, tubes, racks, and gloves were UV-irradiated.
A different pipette tip was used when pipetting each of the PCR reagents, even the same master mix to each tube.
The PCR tubes were kept closed during the procedure. The tubes were opened only when necessary because of potential aerosols that are dangerous with respect to contamination. Minimizing the number of pipetting and mixing steps in PCR master mix preparation is also very important from the perspective of aerosol contamination.
It is very important to schedule PCR when not handling plasmids to prevent cross contamination.
dUTP was incorporated into all PCR products which can subsequently be selectively destroyed by UNG.
Optimization of PCRs is also important. G + C-rich products may be more difficult to inactivate by UNG because of the lower concentration of uridine triphosphate(UTP).
Positive controls consisted of a low copy number of the desired nucleic acid target and should never be prepared or stored with the samples.
Other amplification methods, such as real-time PCR [23] or closed-tube PCR [25] which can reduce the chances of carryover contamination, are now being used more routinely in our laboratory.
A rapid DNA sequencing method to precisely detect contamination was established.
\n
\n
6. Conclusion\n
Standard precautions should always be employed during all PCR-based testing, whether it is real-time PCR or conventional PCR. All the regulatory laboratories should have their own appropriate controls and systematic measures to prevent and detect contamination. When contamination does occur, we need to accurately determine which reagent is contaminated. All of us should also understand that our individual working habits directly affect our work quality. We believe all the above methods can reduce the risk of contamination and ensure the efficacy of all PCR results.
\n
\n
1. Introduction\n
Significant research is being conducted on the development of autonomy for underwater robotic vehicles, which are widely employed in many fields of application such as oceanographic, marine archeology [1], military organizations, and cable tracking and inspection [2]. The advent of underwater robotic vehicles has significantly reduced the dangers in deep sea exploration. Two kinds of robotic vehicles used in marine research are remotely operated vehicles (ROVs) and autonomous underwater vehicles (AUVs). The main difference between the two is that ROVs are connected to the ship by communication cables whereas AUVs operate independently from the ship. AUVs operate without an umbilical; therefore, AUVs are able to conduct a larger range of work area. In this research, AUV intelligent rescue system (AIRS) can minimize the loss of losing or unpredictable catastrophic failure of AUVs caused by emergency conditions.
\n
In recent years, the micro-electro-mechanical system (MEMS) has almost become a vital technology for modern society because of its small volume, low power consumption, low cost, and ease of integration into systems or modification [3]. The MEMS technology creates entirely innovative kinds of products, such as gyroscope sensor in camera-shake detection systems [4], multi-axis inertial motion sensors for smartphone-based navigation [5], and rehabilitation systems based on inertial measurement units (IMUs) [6]. Additionally, the integration of global positioning system and inertial navigation system (GPS/INS) is usually employed to measure the position of AUVs [7, 8]. Unfortunately, small errors in the measurement of initial data are double integrated into larger errors progressively in attitude data, and such errors increase without bound. Error reduction calibration for initialization of INS is paramount for the systematic parameter, like scale factor, bias, and misalignment of the axes. As the situation depicted above, a calibration method is investigated and adopted in this research. There are many researches utilizing Kalman filter (KF) [9], complementary filter (CF) [10], adaptive Kalman filter (AKF) [11], or extended Kalman filter (EKF) [12] to fuse the gyroscope and accelerometer together, taking advantages of their individual strength. In this research, we use EKF to filter IMU outputs with a balance of noise canceling and adaptability simultaneously used in sensing attitude algorithm for AIRS.
\n
The sensing algorithm with EKF for sensing the attitude of AUV, while employing the backpropagation network (BPN) to classify motion data that are formed in the AUV. Before classify the data, there is a problem. The problem is compounded by the fact that our system is computationally complex due to values estimation of AUV and causes the high dimension that is called ''curse of dimensionality,'' which donates the drastic rise of computational complexity and the classification error. Therefore, the step of feature extraction (FE) is an important procedure in data classification [26]. Further, we feed the data into eural network to distinguish every AUV motion in each operating period. The neural network consists of multiple artificial neurons to receive inputs, and process them to obtain an output [13]. By repeating amendments to the model weights, the neural network makes central processing unit (CPU) more logical to calculate the nonlinear systems of AUV motion [14]. The neural network has been widely used because of a number of advantages, including estimating which variables are important in classification, detecting possible interactions between predictor variables, and constructing the prediction model with a high accuracy rate. Furthermore, the neural network has emerged as an important tool in classification which has been investigated in many different important applications [15]. In this study, neural network classifier (NNC) is used in the calculation of nonlinear systems AUV attitude data to detect a catastrophic failure of an AUV. Based on the multilayer feed-forward backpropagation algorithm as NNC, we proposed an effective activity recognition method using 10-DOF sensor module.
\n
The concept of AIRS is inspired from the vehicle safety device: airbags have a flexible fabric bag that can protect and restrict the occupant from being thrown away during a crash accident [16, 17]. Wang and Dragcevic designed the airbags, which fire via a small pyrotechnic charge to increase motorcyclists' safety protections when riding a motorcycle [18, 19]. The airbags are applied to not only fall-protection devices but also automatic inflatable life buoy by the inflatable method of a gas cylinder [20, 21, 22, 23]. We extended the airbag systems to load on the AUV with CO2 cylinder to implement the AIRS. In this research, we developed a malformed detection algorithm for AUV with both EKF and BPN, and the implementation of the airbag system loaded on AUV is underway.
\n
\n
2. Intelligent rescue system\n
In this section, we introduce the hardware components of an AIRS and its working principles. The hardware of AIRS contains the sensing module and a double-trigger inflator. The 11-DOF sensing module combines the 9-DOF inertial sensor, barometric pressure sensor, and temperature sensor to collect the motion data and the depth of the underwater vehicle. The inflator is used with gas cylinders as the double trigger mechanism, which can be triggered not only by electricity but also by water damage when the waterproof cabin is severely broken. Furthermore, we used the commercial software LabVIEW' to analyze the relation between buoyancy force and inflator device in water, to confirm the inflator specifications (Figure 1).
\n
Figure 1.The structure of the intelligent rescue system.
\n
\n
2.1. Sensing module\n
Master control module uses ATmega2560 with the Arduino program written to read motion data of sensor. Ten-axis motion sensor combined with the accelerometer, gyroscope, magnetometer, and the barometer to acquire attitude angle data through the inter-integrated circuit (I2C) is adopted. The communication module uses Bluetooth transmit data wirelessly to the processing software. The block diagram of the 11-DOF sensing module is shown in Figure 2.
\n
Figure 2.The block diagram of the sensing module.
\n
\n
\n
2.2. Inflator mechanism\n
In case of failure and sinking of AUVs underwater, we designed an airbag system to increase the buoyancy and reduce the overall density of the AUVs. AIRS is installed on top of AUV, an airbag inflated in a fault condition to force the AUVs to surface up. In this section, we are going to introduce how to manufacture an inflator mechanism of the AIRS for the AUV, including the specification analysis and the mechanism design.
\n
\n
2.2.1. Specification analysis\n
The AIRS mechanism is installed on top of the AUV and dives down into the ocean. One should note that the hydrostatic pressure exerted by a liquid increases with increase in ocean depth, which influences the specification of the cylinder. Not only that, the full displacement and weight of the AUV also affect the size of the airbag and cylinder. These interrelated factors cause the different performance of the airbag inflation time and rescue time. To analyze the inflator specification more accurately and conveniently, we have engineered the specification calculator for the inflator to ensure that the airbag can be inflated and can bring the AUV to surface successfully. From the required buoyancy and the depth of the location, the calculator analyzes the required gas cylinders and airbag volume.
\n
From the result of the calculation illustrated in Figure 3(b), it takes 68.1 seconds to make nylon airbags fill with CO2 and provides 42.1 N buoyancy for AUV. The design of added buoyancy can bring 7.47 kgw of underwater vehicle to the surface (assume that the full load displacement is fixed). With this technology, we can install the appropriate number of gas cylinders and airbags in accordance with different displacements of AUV accurately and conveniently.
\n
Figure 3.The result of calculation specification analysis. (a) The front panel of the specification calculator. (b) The growth rate between the water depth and the airbag volume.
\n
\n
\n
2.2.2. Mechanism design\n
After the above analysis by the calculator, we found out the appropriate gas cylinder and airbag, and designed the corresponding mechanism parts for the selected gas cylinder and airbag. We are going to elaborate on the design concept of the inflator mechanism in the following. Figure 4 illustrates a double-trigger inflator including a servomotor, horn, spring, spring case, crowbar-striker, water-soluble PVA fiber, release button, and CO2 cylinder. The designed double-trigger inflator can be triggered not only by electricity but also by water damage when the waterproof cabin is severely broken.
\n
After the above analysis by the calculator, we found out the appropriate gas cylinder and airbag, and design the corresponding mechanism parts for the selected gas cylinder and airbag.
\n
Figure 4.The design diagram of a double-trigger inflator. (a) before release. (b) release by servo motor. (c) release by spring.
\n
There are two operating mechanisms to trigger the inflator. To begin with the electricity trigger mechanism as shown in Figure 4(b), we assume that the waterproof cabin is watertight; after detecting the failure of AUVs, servo motor rotates for pushing the crowbar-striker, and the crowbar-striker punches the release button, which needs 35 N normal force to release into the airbag through the air tube. We chose MG995 for servomotor, which has product size 40.7 19.7 42.9 mm and weight 55 g; MG995 is sufficient to move the horn for the release button with its maximum torque 127 N-cm.
\n
Besides, if the AUV is severely impacted and causes cabin seepage, so that the power system is damaged, the AUV cannot resurface by the thruster or the electricity-trigger inflator. Hence, we designed the water-trigger inflator for this kind of case. When the cabin is broken, water melts the water-soluble PVA fiber and releases the spring case as shown in Figure 4(c). Within the spring case, the music wire is selected to spring; the true maximum load of spring is 27.985 N, which is not enough to press the release button. Thus, we extended the striker into a crowbar-striker, and amplified the spring force by the lever principle to release the push button. The final part is the airbag, which is made of nylon 6/6 and folded under a cover. The airbag weighs 150 g and the value of tensile strength is 809.3 atm, which can withstand the pressure caused by gas cylinders.
\n
\n
\n
\n
3. Sensor with activity recognition algorithm\n
The method and the procedure of data processing and activity recognition are described in Section 3. The former includes the signal calibration and extended Kalman filter algorithm; the latter is constructed of FE and BPN classifier.
\n
\n
3.1. Data processing of 11-DOF sensing module\n
The 10-DOF sensing module including the 9-DOF IMU (three accelerometers, three gyroscopes, three magnetometers), barometer, and thermometer integrated with a microcontroller (Arduino Mega2560). The six-axis inertial sensor (MPU6050), which is a complete triple-axis gyroscope and triple-axis accelerometer inertial sensing system, is the most suitable sensor for stabilization and attitude measurement. MPU6050 contains the digital motion processor (DMP) which performs the motion processing algorithm itself. However, the horizontal attitude (yaw) is not a possible measurement. We used a magnetometer (HMC5883L) with fusion algorithm accordingly for eliminating the gyroscope offset to recognize the AUV activity context more reliably. The barometer (BMP180) is used in AIRS to detect the depth of the AUV and control the pressure of the airbag for supplying the perfect buoyancy. In addition, BMP180 also includes a temperature sensor, which does not only measure the underwater temperature but also thermal compensation for the MPU6050. Although the development of the MEMS technology has made a great progress, the IMUs are still difficult to collect precise data in the presence of various errors. Calibration is the process of comparing the measurement outputs with known reference data and determining the coefficients that drive the output to agree with the reference data over a range of output. We calibrated the sensor in terms of its thermal noise, bias, and scale factor.
\n
\n
\n
3.2. EKF strategy\n
A quaternion-based EKF is proposed in this section for determining the attitude of the AUV from the outputs of IMU. Attitude estimation is a very important part of the AIRS. If the initial data are double integrated into larger errors without bound, it brings on the misclassification of ANN algorithm with wrong attitude values. For the sake of this error, we used a series of measurements and observed over time for signal processing. The main advantage of the KF is its ability to provide the quality of the estimate, but the KF only applies to linear and Gaussian models. The EKF conversely is a nonlinear version of the KF which linearizes about an estimate of the current mean and covariance. In view of this, we chose the EKF to filter IMU outputs with a balance of noise canceling and adaptability simultaneously, used in sensing attitude algorithm for AIRS. We proposed the EKF fusing with the accelerometer, gyroscope, and magnetometer integrated with sensor calibration (SC). SCEKF results in an improvement of the orientation accuracy from IMU. A flow-chart of structure performed by the proposed SCEKF is capsuled in Figure 5.
\n
Figure 5.Overview of the SCEKF structure.
\n
\n
3.2.1. Prediction step\n
\n\n\n\nx\nÌ\n\nk\n\n=\nÎ...\n\n\nT\ns\n\n\n\nÏ\nÌ\n\n\nk\n''\n1\n\n\n\n\nx\n\nk\n''\n1\n\n\n2\n\n\n+\n\n\nν\nÌ\n\n\nk\n''\n1\n\n\n\n E1
\n
\n\n\n\nP\nÌ\n\nk\n\n=\n\nF\n\nk\n''\n1\n\n\n\nP\n\nk\n''\n1\n\n\n2\n\n\n\n\nF\n\nk\n''\n1\n\n\nT\n\n+\n\nB\n\nk\n''\n1\n\n\n\nQB\n\nk\n''\n1\n\nT\n\n\n E2
\n
The first step of EKF predicts a current state and covariance matrix at time \n\nk\n\n . We estimated a current state based on the previous states \n\n\n\nx\nÌ\n\n\nk\n''\n1\n\n\n=\n\n\n\n\n\na\nÌ\n\n\nk\n''\n1\n\n\n\n\nÏ\nÌ\n\n\nk\n''\n1\n\n\n\n\nm\nÌ\n\n\nk\n''\n1\n\n\n\n\nT\n\n\n that are composed of the gyroscope measurement \n\n\n\nÏ\nÌ\n\n\nk\n''\n1\n\n\n\n and white noise \n\n\n\nν\nÌ\n\n\nk\n''\n1\n\n\n=\n\n\n\n\n\n\nν\nÌ\n\n\n\na\n\n\nk\n''\n1\n\n\n\n\n\nν\nÌ\n\n\n\nÏ\n\n\nk\n''\n1\n\n\n\n\n\nν\nÌ\n\n\n\nm\n\n\nk\n''\n1\n\n\n\n\nT\n\n\n and the priori covariance matrix \n\n\n\nP\nÌ\n\nk\n\n\n based on a previous covariance matrix \n\n\nP\n\nk\n''\n1\n\n\n2\n\n\n\n , covariance \n\nQ\n\n and a state noise coefficient matrix \n\n\nB\n\nk\n''\n1\n\n\n\n .
\n
\n
\n
3.2.2. Jacobian computation\n
\n\n\nF\n\nk\n''\n1\n\n\n=\n\n'Î...\n\n'\nx\n\n\n\n\n\n\n\n\nT\ns\n\n,\n\n\nÏ\nÌ\n\n\nk\n''\n1\n\n\n\n\n\n=\n\nJ\nf\n\n\n\nT\ns\n\n\n\nÏ\nÌ\n\n\nk\n''\n1\n\n\n\n\n E3
\n
\n\n\n\ny\nË'\n\nk\n\n=\n\nz\nk\n\n''\nh\n\n\n\nx\nÌ\n\nk\n\n0\n\n,\n\n\nH\nk\n\n=\n\n\n'\nh\n\n\n'\nx\n\n\n\n\n\n\n\n\nx\nÌ\n\nk\n\n\n\n=\n\nJ\nh\n\n\n\n\nx\nÌ\n\nk\n\n\n\n E4
\n
However, in view of the nonlinear process of state transition \n\n\nF\n\nk\n''\n1\n\n\n\n and observation \n\n\nH\nk\n\n\n directly, the EKF approach requires estimation by computing the Jacobian.
\n
\n
\n
3.2.3. Gravity corrections\n\n
\n\n\nS\nk\n\n=\n\nH\nk\n\n\n\nP\nÌ\n\nk\n\n\n\nH\nk\n\nT\n\n+\nR\n\n\n\n\na\n\n\n\n E5
\n
\n\n\nK\nk\n\n=\n\n\nP\nÌ\n\nk\n\n\n\nH\nk\n\nT\n\n\n\n\n\nH\nk\n\n\n\nP\nÌ\n\nk\n\n\n\nH\nk\n\nT\n\n+\nR\n\n\n\n\na\n\n\n\n\n\n''\n1\n\n\n\n E6
\n
\n\n\nx\nk\n\n1\n\n\n=\n\n\nx\nÌ\n\nk\n\n+\n\nK\nk\n\n\n\n\na\nk\n\n''\n\nH\nk\n\n\n\nx\nÌ\n\nk\n\n\n\n\n E7
\n
\n\n\nP\nk\n\n1\n\n\n=\n\n\n\nI\n\n4\n\n4\n\n\n''\n\nK\nk\n\n\nH\nk\n\n\n\n\n\nP\nÌ\n\nk\n\n\n E8
\n
where the innovation covariance matrix \n\n\nS\nk\n\n\n is based on the a priori error covariance matrix \n\n\n\nP\nÌ\n\nk\n\n\n and the measurement covariance matrix of accelerometer \n\nR\n\n\n\n\na\n\n\n\n , whose main diagonal elements are from the accelerometer values, and nonmain diagonal elements are all zero conversely. The Kalman gain \n\n\nK\nk\n\n\n is the error covariance matrix after gravity measurement is corrected.
\n
\n
\n
3.2.4. Yaw corrections\n
\n\n\nS\nk\n\n=\n\nH\nk\n\n\nP\nk\n\n1\n\n\n\nH\nk\nT\n\n+\n\nR\nm\n\n\n E9
\n
\n\n\nK\nk\n\n=\n\nP\nk\n\n1\n\n\n\n\nH\nk\n\nT\n\n\n\n\n\nH\nk\n\n\nP\nk\n\n1\n\n\n\n\nH\nk\n\nT\n\n+\n\nR\nm\n\n\n\n\n''\n1\n\n\n\n E10
\n
\n\n\nx\nk\n\n2\n\n\n=\n\nx\nk\n\n1\n\n\n+\n\nK\nk\n\n\n\n\nm\nk\n\n''\n\nH\nk\n\n\nx\nk\n\n1\n\n\n\n\n\n E11
\n
\n\n\nP\nk\n\n2\n\n\n=\n\n\n\nI\n\n4\n\n4\n\n\n''\n\nS\nk\n\n\nH\nk\n\n\n\n\nP\nk\n\n1\n\n\n\n E12
\n
where the measurement covariance matrix of magnetometers \n\nR\n\n\n\n\nm\n\n\n\n , which main diagonal elements are from the magnetometer values, non-main diagonal elements are all zero conversely. \n\n\nP\nk\n\n2\n\n\n\n is the error covariance matrix after yaw measurement is corrected.
\n
\n
\n
\n
3.3. Feature extraction (FE)\n
AUV motions are defined by the six degrees of freedom, including heave, surge, sway, pitch, roll, and yaw, and they are coupling by the vehicle shape, trends, and current interaction. Therefore, the attitude data of AUV is of high dimension and very complex. The value of FE is to reduce the dimension of the large measurement data and prevent program operation to run out of memories. The characteristics of a data segment are to keep the most meaningful features and remove the redundant data. Therefore, the FE methods have been applied for activity detection from accelerometer data [24, 25]. İn order to extract features easily, the continuous measurement data of sensors are divided into many overlapping segments of which each is 20 seconds long, as illustrated in Figure 6. A sliding window technique cuts the sensors' data into 20 seconds (550 samples) in each short-term window with 50% overlapping. As previous studies found that the effectiveness of FE on windows with 50% overlap is an effective window size [24].
\n
Figure 6.Attitude data with a long-term window, which consists of many overlapping short-term windows.
\n
In this chapter, we adopted principal component analysis (PCA) as the feature selection procedure to lower the dimension of the original features. Feature extraction is highly subjective in nature, it depends on applications. Here, we introduced the following features that are beneficial to the classification of AUV failure detection, and used these features to discriminate the type of AUV activity: (1) min, max; (2) mean; (3) interquartile range (IQR); (4) root mean square (RMS); (5) standard deviation (STD); (6) root mean square error (RMSE); (7) signal magnitude area (SMA); (8) signal vector magnitude (SVM); and (9) averaged acceleration energy (AEE).
\n
\n
\n
3.4. BPN classifier\n
Finally, this feature extraction by the PCA is fed into the short-term classifier and long-term classifier sequentially, and then output the AUV conditions. In order to detect the AUV condition accurately, reliably, stably, and robustly, we divided the recognition system into three classifiers which are based on BPN. BPN is considered the workhorse of ANNs and is the multilayer perceptron (MLP) based on a feed-forward algorithm. The hidden layers, between input and output layers, use the error backpropagation (BP) algorithm to compute the nonlinear relationship in supervised learning as shown in Figure 7. The main features of BPN are as follows: high learning accuracy, fast response, and ability to process the nonlinear problems.
\n
Figure 7.The topology of the BPN classifier.
\n
In this chapter, a three-layer BPN is used for classifying AUV failure conditions. The topology of the BPN classifier is shown in Figure 7 [26]. The input layer has R neurons, equal to the dimension of the feature vectors \n\nI\n=\n\n\n\nI\n1\n\n1\n\n\n\n\n\nI\nr\n\n1\n\n\n\n'...\n\nI\nR\n\n1\n\n\n\nT\n\n\n , where superscript 1 indicates the first layer. The hidden layer has J neurons, and the output layer has P neurons, equal to the number of AUV conditions \n\nO\n=\n\n\n\nO\n1\n\n3\n\n\n\n\n\nO\np\n\n3\n\n\n\n'...\n\nO\nP\n\n3\n\n\n\nT\n\n\n , where superscript 3 indicates the third layer. \n\n\nw\npj\n\n3\n\n\n\n and \n\n\nw\njr\n\n2\n\n\n\n denote the weight from the hidden to the output layer and from the input to the hidden layer, respectively. \n\n\nb\n\n3\n\n\n\n and \n\n\nb\n\n2\n\n\n\n denote the bias in the third and second layers, respectively. \n\n\ni\n\nn\n\n\n\n and \n\n\no\n\nn\n\n\n\n are the input value and the activation of a neuron in the nth layer \n\n\n\nn\n'\nN\n:\n\nn\n\n'¤\n3\n\n\n\n . We take the interval between the hidden layer (second) and the output layer (third) as an example. The function of the neurons in each step is defined as follows [27]:
\n
In the BPN, the intermediate quantity \n\n\nnet\nj\n\n3\n\n\n\n is the weighted input to the neurons in the third layer, and implements a nonlinear transformation from the output values of the second layer to the output values of the third layer. \n\nÏ\n\n is called the sigmoid function and is in general nonlinear and differentiable. The functions of the pth neuron are given by:
\n
\n\n\no\np\n\n3\n\n\n=\nÏ\n\n\nnet\nj\n\n3\n\n\n\n=\nÏ\n\n\n\n'\n\nj\n=\n1\n\nJ\n\n\nw\npj\n\n3\n\n\n\no\nj\n\n2\n\n\n+\n\nb\np\n\n3\n\n\n\n\n\n E13
\n
The discrepancy \n\n\nE\n\n3\n\n\n\n between the desired output \n\n\nt\np\n\n3\n\n\n\n and the real output \n\n\nO\np\n\n3\n\n\n\n in the third layer can motivate the BP learning algorithm as gradient descent on the sum-squared discrepancy. The factor of \n\n1\n/\n2\n\n will simplify the exponent when differentiating later. The function is defined as:
\n
\n\n\nE\n\n3\n\n\n=\n\n1\n2\n\n\n'\np\n\n\n\n\n\nt\np\n\n3\n\n\n''\n\nO\np\n\n3\n\n\n\n\n2\n\n\n E14
\n
Then, the weights are adjusted to find the partial derivative \n\nE\n\n with respect to a weight \n\n\nw\npj\n\n3\n\n\n\n : \n\nÎ--\n\nw\npj\n\n3\n\n\n''\n''\nη\n\n\n'\n\nE\n\n3\n\n\n/\n'\n\nw\npj\n\n3\n\n\n\n\n\n . However, the discrepancy is not directly the function of weights. Calculating the derivative of the discrepancy is done using the chain rule twice.
\n
\n\nÎ--\n\nw\npj\n\n3\n\n\n=\n''\nη\n\n\n'\n\nE\n\n3\n\n\n\n\n'\n\no\np\n\n3\n\n\n\n\n\n\n'\n\no\np\n\n3\n\n\n\n\n'\n\nnet\np\n\n3\n\n\n\n\n\n\n'\n\nnet\np\n\n3\n\n\n\n\n'\n\nw\npj\n\n3\n\n\n\n\n\n E15
\n
Let us discuss each of partial derivatives in turn. In the last term of Eq. (15) is the derivative of the net \n\n\nnet\np\n\n3\n\n\n\n with respect to a weight \n\n\nw\npj\n\n3\n\n\n\n :
\n
\n\n\n\n'\n\nnet\np\n\n3\n\n\n\n\n'\n\nw\npj\n\n3\n\n\n\n\n=\n\n'\n\n'\n\nw\npj\n\n3\n\n\n\n\n\n\n\n'\n\nj\n=\n1\n\nJ\n\n\nw\npj\n\n3\n\n\n\no\nj\n\n2\n\n\n+\n\nb\np\n\n3\n\n\n\n\n=\n\no\nj\n\n2\n\n\n\n E16
\n
Next, the derivative of the activation \n\n\no\np\n\n3\n\n\n\n with respect to the net input \n\n\nnet\np\n\n3\n\n\n\n :
\n
\n\n\n\n\n\n\n'\n\no\np\n\n3\n\n\n\n\n'\n\nnet\np\n\n3\n\n\n\n\n=\n\n'\n\n'\n\nnet\np\n\n3\n\n\n\n\nÏ\n\n\nnet\np\n\n3\n\n\n\n\n\n\n\n\n=\nÏ\n\n\nnet\np\n\n3\n\n\n\n\n\n1\n''\nÏ\n\n\nnet\np\n\n3\n\n\n\n\n\n=\n\no\np\n\n3\n\n\n\n\n1\n''\n\no\np\n\n3\n\n\n\n\n\n\n\n\n E17
\n
Last, we consider the derivative of \n\nE\n\n with respect to the activation. Since the neuron is in the output layer, then \n\n\nO\np\n\n3\n\n\n\n is equal to \n\n\no\np\n\n3\n\n\n\n .
\n
\n\n\n\n'\nE\n\n\n'\n\no\np\n\n3\n\n\n\n\n=\n\n\n'\nE\n\n\n'\n\nO\np\n\n3\n\n\n\n\n=\n\n\n'\n\n\n1\n/\n2\n\n\n\n\nt\np\n\n3\n\n\n''\n\nO\np\n\n3\n\n\n\n\n2\n\n\n\n\n\n'\n\nO\np\n\n3\n\n\n\n\n=\n''\n\n\n\nt\np\n\n3\n\n\n''\n\no\np\n\n3\n\n\n\n\n\n E18
\n
Finally, we substitute these results Eqs. (15)''(17) back into original Eq. (17) to find the weight change \n\nÎ--\n\nw\npj\n\n3\n\n\n\n rule,
\n
\n\nÎ--\n\nw\npj\n\n3\n\n\n=\nη\n\n\n\n\n\n\n\n\nt\np\n\n3\n\n\n''\n\no\np\n\n3\n\n\n\n\n\no\np\n\n3\n\n\n\n\n1\n''\n\no\np\n\n3\n\n\n\n\n\n¸·\n\n\n\nε\np\n\n\n\n\no\nj\n\n3\n\n\n=\n\nηε\np\n\n\no\nj\n\n3\n\n\n\n E19
\n
We can find the weight change \n\nÎ--\n\nw\njr\n\n2\n\n\n\n in the same processes as above and it is defined as:
\n
\n\nÎ--\n\nw\njr\n\n2\n\n\n=\nη\n\n\n\n\n\n\n'\nj\nJ\n\n\nε\np\n\n\nw\npj\n\n3\n\n\n\no\nj\n\n2\n\n\n\n\n1\n''\n\no\nj\n\n2\n\n\n\n\n\n¸·\n\n\n\nε\nj\n\n\n\n\no\nr\n\n1\n\n\n=\n\nηε\nj\n\n\no\nr\n\n1\n\n\n\n E20
\n
The BPN algorithm approach to recognize and intelligently detect failures is based on changes in weight values \n\nÎ--\n\nw\npj\n\n3\n\n\n\n \n\nÎ--\n\nw\njr\n\n2\n\n\n\n of the AUV feature parameters.
\n
\n
\n
4. Experimental test and result\n
\n
4.1. SCEKF\n
To validate the sensor calibration and fusion method proposed in this chapter, this section covers the results obtained from the 10-DOF sensor module with operating system of Windows 8.1, 2.20 GHz CPU, 8 GB memory, and experiments are performed with the Arduino program described in Section 2.1. The system was fixed on an anti-vibration table to minimize interference. The proposed SCEKF algorithm was used to estimate accelerometer and orientation with 27-Hz updating rate.
\n
As the calculation process of sensor algorithm, we used the rotary platform with outputting the quantitative and stable signal to observe the relative signal output. The first process is fixing IMU at the center and along the rotation axis of the platform and fixing it by valves. After installation, setting the required rate or angle of rotation as a reference data by the user interface, we measured the output corresponding to the reference data. The performance of the orientation before/after SCEKF algorithm is presented in Figure 8(a''c), respectively; it produced reasonable output values within the expected ranges. It is seen that each component of Euler orientation (i.e., roll, pitch, and yaw) is within 0·1 degree after SCEKF processing in the static test.
\n
Figure 8.The calibration results from the static condition produced by the SCEKF processing. (a) Roll angle calibration.(b) Pitch angle calibration. (c) Yaw angle calibration.
\n
\n
\n
4.2. AUVs' attitude simulation and experiment\n
\nFigure 9 shows the whole system of the TrenchRoDer equipped with the proposed AIRS. The whole equipment includes the TrenchRoDer, a camera, an AIRS, a Bluetooth box, and a floating board. We put a camera in the waterproof bags and tied them to a fixed plate in order to observe the situation when the airbag is inflated. The function of the floating board is to protect the Bluetooth box from immersing into the water.
\n
Figure 9.The TrenchRoDer equipped with the proposed AIRS.
\n
In order to construct a robust underwater vehicle fault attitude database, we simulated the underwater vehicle motion model with different situations in the water environment, which can provide the particular case of attitude data, such as if the AUVs suffered a crash or a disabled propeller. Moreover, we built the fault simulator GUI interface for a more convenient operation, in other words, the AIRS starts with the fault simulator GUI interface, after inputting the required AUV's dimensions, hydrodynamic coefficients, buoyancy center, and inertia coefficients. Next, the simulator calculates the different AUV motion data under different cases through the Matlab' program as shown in Figure 10. The above results from simulator will combine with the experimental data as the database for the training and verification of BPN classifier. Last but not least, the real AUV motion signal is set as the testing dataset to ensure the establishment of the AIRS. The modular modeling equation of AUVs is selected from [28]. The modeling method of Prestero's which id without considering the sea conditions of emission can help us to generate the dynamic models of AUVs quickly and conveniently. The dynamics model and kinematics model of AUVs are established by analysis of the force working on AUVs moving underwater, based on the theorem of the momentum of the rotation around the buoyancy center and the theorem of the motion of mass center. The motion in 6-DOF of the AUVs is determined. And then, the attitude data of AUVs at any instant are determined.
\n
Figure 10.The AUV's fault simulator diagram.
\n
Since we simulated the AUV motion in the underwater environment without considering the influence of currents and waves, we did the wave maker experiment in NCKU's ship model towing tank to observe the changes and effects on the AUV motion data for different wave heights. By experimenting with simulation, we can be closer to the state of the real ocean environment. The experiment of the AUVs affected by wave maker in NCKU's ship model towing tank is illustrated in Figure 11.
\n
Figure 11.The wave maker of the NCKU's ship model towing tank.
\n
\n
\n
4.3. Classification results\n
In this chapter, we focused on 10 conditions of AUV, which are listed in Table 1.
\n
\n\n\n\n\nNumber\nActivity description\n\n\n\n\n1\nFunctional condition after horizontal sinking\n\n\n2\nMalfunction after horizontal sinking\n\n\n3\nFunctional condition after vertical sinking\n\n\n4\nMalfunction after vertical sinking\n\n\n5\nFunctional condition after ramped sinking\n\n\n6\nMalfunction after ramped sinking\n\n\n7\nFunctional condition after turbulence\n\n\n8\nMalfunction after turbulence\n\n\n9\nFunctional condition after collision\n\n\n10\nMalfunction after collision\n\n\nTable 1.Activities performed in this experiment.
\n
There are two types of these AUV conditions. One of them is failure situation and the other is functional condition and both include five motion status of the AUV. We carried out 20 experiments for capturing data and verifying classifier. The data from 17 experiments were adopted in the training program of the recognition scheme; the data obtained from the other experiments were used for testing the recognition performance. Note that, since the sampling frequency is 27.5 Hz, the total number of the short-term and long-term samplings for each activity of each experiment is 550 and 2200, respectively, which means 20 seconds per short-term window and 80 seconds per long-term window. The feature extraction of this chapter was based on 50% overlapping windows using 550 samples of window sizes to avoid information loss at the boundary of a single window. The dimension of a feature vector was 45 (an accelerometer 3 axes 9 features + a gyroscope 3 axes 6 features). Figure 12 illustrates the first 2200 data of accelerations and Euler orientations collected from the first experiment. The selected features of sensor's data enabled effective recognition of the conditions and were suggested for BPN training procedure. A computation program adopted the input features and activated the feature classifier learning procedure with the BP algorithm, and outputted the results to short-term classifier. Then, an AUV condition was distinguished by a long-term classifier, of which the input is from the short-term classifier to raise the accuracy of failure recognition. The number of neurons in each hidden layer is 4, 6, and 7 for the feature classifier, short-term classifier, and long-term classifier, respectively, and the number of epochs is 700 for each neural training. The BPN classifier was trained on the training data set and tested on the test set which are from the experiment values. The classifier was created by neural network toolbox of MATLABTM for practical implementation and for validating the proposed model.
\n
Figure 12.The accelerations and Euler orientation of the first experiment.
\n
After building up our prediction algorithm, we apply our chosen prediction algorithm on our new test set which is from the real signal of AUV, in order to have an idea about the algorithm's performance on unseen data. The confusion matrix measured in the real AUV test is shown in Table 2, which recorded the results from 20 times experiments on each condition of AUV. We have implemented in two different ways under MATLABTM environment. In the first, we conducted in our proposed classifier system with feature extraction and the results indicate that the AUV failure detection on the average 97% of the time, and a successful functional condition accuracy of 93% is achieved. Second, we chose a classifier in ANN learning algorithm without feature extraction for comparing with the classifier that we proposed. The performance indicated that the ANN classifier without feature extraction performs poorer than our proposed classifier. From the confusion matrix, we can know that the malfunction and functional conditions are not easy to be confused. However, the motions within functional condition may be misclassified between each other, because these activities contain similar amplitude peaks and waveforms at the AUV.
\n
\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n20 times of testing experiments\n\n\nClassified Type\n1\n2\n3\n4\n5\n6\n7\n8\n9\n10\nRecognition rate (%)\nDetection accuracy (%)\n\n\n\n\n1\n14/13\n0/0\n1/2\n0/0\n5/6\n0/0\n0/0\n0/0\n0/0\n0/1\n70/65\nMalfunction\n97/92\n\n\n2\n1/2\n16/14\n0/0\n0/0\n0/0\n1/2\n0/0\n0/1\n2/3\n2/4\n80/70\nFunctional condition\n93/88\n\n\n3\n1/1\n0/0\n15/13\n0/0\n0/1\n0/0\n0/0\n0/0\n1/1\n0/0\n75/65\n\n\n\n\n4\n0/0\n0/0\n0/0\n18/16\n0/0\n2/2\n0/1\n0/0\n0/0\n0/0\n90/80\n\n\n\n\n5\n4/3\n0/0\n2/4\n0/1\n15/13\n0/0\n0/0\n0/0\n0/1\n0/0\n75/65\n\n\n\n\n6\n0/1\n1/1\n0/0\n2/3\n0/0\n16/15\n0/0\n3/4\n0/0\n0/0\n80/75\n\n\n\n\n7\n0/0\n0/0\n0/0\n0/0\n0/0\n0/0\n17/15\n1/2\n0/0\n0/0\n85/75\n\n\n\n\n8\n0/0\n0/0\n0/0\n0/0\n0/0\n1/1\n3/4\n16/13\n0/0\n0/0\n80/65\n\n\n\n\n9\n0/0\n1/2\n2/1\n0/0\n0/2\n0/0\n0/0\n0/0\n16/14\n1/2\n80/70\n\n\n\n\n10\n0/0\n2/3\n0/0\n0/0\n0/0\n0/0\n0/0\n0/0\n1/1\n17/14\n85/70\n\n\n\n\nTable 2.Confusion matrix for all the testing experiments.
\n* Proposed classifier with feature extraction in BP algorithm/ANN classifier without feature extraction.
\n
\n
\n
5. Conclusions and future works\n
The main objective of this work was to develop an intelligent AUV rescue system (AIRS) for detecting failure to minimize loss of an autonomous underwater vehicle (AUV). We combined three main ideas to construct the AIRS, including sensors algorithm, classifier conducting, and airbag system. Complex data acquisition was done by the 10-DOF sensor module with sensor calibration and extended Kalman filter (SCEKF), where Euler orientation fused with gravity and magnetic field are state variables, can benefit to get the precise attitude from the AUV. After SCEKF processed, we extracted the features of these signals from the 10-DOF sensor module and selected these features by principal component analysis (PCA) method. The results were incorporated with feature classifier, short-term classifier, and long-term classifier in order to recognize 10 types of AUV conditions. According to the experimental test in Section 3, we have shown that the 20 experimental data sets are categorized into ''malfunction'' or ''functional condition'' categories. The outcomes of the proposed classification with features extracted, whose failure detection accuracy is 97%, were more accurate than those of the ANN without features extracted. These results confirmed that the technology of the AIRS was feasible and that the proposed methods were accurate. Furthermore, we designed an inflatable mechanism, which fills CO2 in the airbags to generate buoyancy for AUV during failure detection.
\n
The attitude estimation and classification applied in the underwater environment are a new field. Considering future work, we will try to extend more condition types of AUV for classifying more complex situations and accomplish the airbag system for setting on the AUV. We can have the ability to construct a variety of different models for the AUV's fault simulation, such as underwater turbulence or underwater creatures' interference. Providing a more comprehensive AIRS, to not have to do the experiment in the real underwater environment, also can improve the performance and convenience of installation to the AIRS. With this technology, we can install the appropriate number of gas cylinders and airbags in accordance with the different displacements of AUV to avoid the loss of AUV; this can even be used in rescuing vessels to reduce shipwreck in the future, thereby minimizing loss of life and property. This study will have outstanding contributions for the next generation of underwater vehicles. We are looking forward to the application of the SIRS being used widely in the future.
\n
Acknowledgments\n
The authors hereby extend sincere thanks to the Ministry of Science and Technology (MOST) and Fisheries Agency for their financial support of this research, whose project codes are MOST 106-2221-E-006-121, MOST 107-2218-E-006-031, MOST 107-2218-E-110-004-and 107AS-14.2.7-FA-F1(3) and 107è¾²ç§-14.2.7-æ¼-F1(3). We thank to the generous patronage of the MOST and Fisheries Agency that this study has been smoothly performed.
\n
You have been successfully unsubscribed.